a

r
X
i
v
:
q
u
a
n
t
-
p
h
/
0
4
0
4
0
1
7
v
2


1
3

M
a
y

2
0
0
4
Quantum Mechanical Properties of Bessel Beams
R. Jauregui and S. Hacyan
Instituto de Fsica, U.N.A.M., Apdo. Postal 20-364, Mexico D. F. 01000, Mexico.
(Dated: February 1, 2008)
Abstract
Bessel beams are studied within the general framework of quantum optics. The two modes of the
electromagnetic eld are quantized and the basic dynamical operators are identied. The algebra
of these operators is analyzed in detail; it is shown that the operators that are usually associated to
linear momentum, orbital angular momentum and spin do not satisfy the algebra of the translation
and rotation group, and that this algebra is not closed. Some physical consequences of these results
are examined.
PACS numbers: 42.50.Vk, 32.80.Lg
1
I. INTRODUCTION
The increasing use of light to control the motion of atomic systems and microparticles
has renewed the interest in the mechanical properties of electromagnetic (EM) beams. The
realization that light carries energy as well as linear and angular momenta was essential
in the development of the classical theory of electromagnetism; nowadays, the interchange
of such quantities with matter is a well established fact. In most cases, the interaction of
light with matter can be satisfactorily described by taking the EM eld as a superposition
of idealized plane waves; in this simple picture, each normal mode carries momenta in the
direction of the propagation vector, and the angular momentum is directly related to the
states of polarization.
However, the decomposition of angular momentum into an orbital and spin part has some
ambiguities within the framework of quantum optics, as was recognized in classical papers
by Darwin [1], Humblet [2], de Broglie [3], and many others. Recent interest in dening
angular momentum in optics can be traced back to the works of Lenstra and Mandel [4],
Allen et al [5], and van Enk and Nienhuis [6, 7]. Lenstra and Mandel considered periodic
boundary conditions that limit the isotropy of the EM eld and thus aect the angular
momentum. Allen et al showed that angular momentum can indeed be decomposed into
orbital and spin parts for Laguerre-Gaussian modes, i. e., paraxial elementary waves with
cylindrical symmetry. Finally, van Enk and Nienhuis [7] studied this decomposition and
showed that neither the spin nor the orbital quantum operators of the electromagnetic eld
satisfy the commutation relations of angular momentum; they made the assumption that
the total angular momentum obtained from Noether theorem is given directly as a sum of
these operators; however, this is not the case in many situations where boundary conditions
give rise to important surface eects.
In general, orbital angular momentum (OAM) is identied with the part of the total
angular momentum that depends explicitly on position (with respect to an origin of coor-
dinates). When OAM is evaluated with respect to the axis of symmetry, it turns out that
Laguerre-Gaussian modes do carry OAM in integer multiples of [5]. However, the formal
decomposition into OAM and spin angular momentum does not appear to be natural beyond
the paraxial approximation [8]. Nevertheless, recent experiments have shown that angular
momentum do indeed possess orbital and spin parts [10, 11, 12, 13, 14], the former having
2
an intrinsic and extrinsic nature with direct physical consequences [15].
Bessel beams have interesting properties that make them especially attractive: they prop-
agate with an intensity pattern invariant along a given axis [16] and carry angular momentum
along that axis. Experimental realizations of such beams and their mechanical eects on
atoms and microparticles are the subject of many current investigations [13]. The purpose
of the present paper is to study the mechanical properties of Bessel beams within the general
formalism of quantum optics. In particular, we study the standard decomposition of the
total angular momentum of these beams into orbital and spin parts, and obtain explicit
expressions for these observables as quantum operators. It turns out, however, that they
do not satisfy the usual commutation relations for angular momentum vectors and that,
contrary to van Enk and Nienhuis [7] results, the algebra does not even close; we argue that
this discrepancy is due to the boundary conditions. We also show that the superposition of
Bessel modes, as well as their polarization states, can be characterized by a set of opera-
tors that appears naturally within our formulation and can be measured in principle. Some
predictions that could eventually be tested experimentally are discussed in the conclusions.
The organization of the article is as follows. In section II the electromagnetic Bessel
beams are expressed in terms of Hertz potentials. In Section III the elds are quantized
and explicit expressions for the operators are given, relating them to their main mechanical
properties. Section IV is devoted to the study of the algebraic properties of the operators that
are usually identied with orbital and spin angular momentum. Finally, a brief discussion
of the results and their experimental implications is presented in Section V. Some useful
formulas are given in Appendix A, and expressions relating Bessel modes with plane and
spherical EM modes are given in Appendix B.
II. ELECTROMAGNETIC BESSEL MODES
An electromagnetic eld with cylindrical symmetry can be conveniently described in the
terms of Hertz potentials
1
and
2
[17]. In cylindrical coordinates {, , z}, the electro-
magnetic potentials are given by
=

z

1
, (1)
3
A =
_
1

2
,

2
,

ct

1
_
, (2)
and satisfy the Lorentz gauge condition. Then, the electric eld E is:
E

=

2
z

2
ct

2
, (3)
E

=
1

2
z

1
+

2
ct

2
, (4)
E
z
=

2
c
2
t
2

1
+

2
z
2

1
, (5)
and the magnetic eld B is:
B

=
1

2
ct

1
+

2
z

2
, (6)
B

=

2
ct

1
+
1

2
z

2
, (7)
B
z
=

2
c
2
t
2

2
+

2
z
2

2
. (8)
Both Hertz potentials
i
, (i = 1, 2), satisfy the equations:

2
c
2
t
2

i
+
1

i
_
+
1

i
+

2
z
2

i
= 0. (9)
Any solution of this equation that is regular at the origin can be written as a linear
combination of the functions

i
= C
i
J
m
(k

) exp{it + ik
z
z + im}, (10)
where J
m
is the Bessel function of order m, C
i
are constants, and k

=
_
(/c)
2
k
2
z
. Bessel
functions form a complete orthogonal basis as follows from Eq. (73) in Appendix A.
An electromagnetic mode is associated to each Hertz potential,
1
and
2
, via Eqs. (5-8),
giving rise to transverse magnetic and electric modes respectively. In the following we will
occasionally denote them by the superscripts TM and TE.
It is convenient to make a further gauge transformation to the transverse gauge with
= 0. This can be achieved with the transformations /ct and A A+,
and taking = (k
z
c/)
1
.
4
The electromagnetic vectors and potential can be decomposed in terms of their basic
modes. First, we dene the vectors:
M(r, t; K) =

ck
z
_
m
k

J
m
(k

)e

+ iJ

m
(k

)e

_
e
it+im+ikzz
(11)
and
N(r, t; K) =
_
iJ

m
(k

)e

m
k

J
m
(k

)e

+
k

k
z
J
m
(k

)e
z
_
e
it+im+ikzz
. (12)
Here and in the following, the set of quantum numbers {k

, m, k
z
} are denoted by the generic
symbol K whenever no confusion can arise.
Accordingly, we obtain the following forms for the modes of the electromagnetic potential:
A
(TM)
(r, t; K) =
c
i
E
(TM)
m
(k

, k
z
)N(r, t; K) (13)
and
A
(TE)
(r, t; K) =
c
i
E
(TE)
m
(k

, k
z
)M(r, t; K), (14)
where the functions E
(i)
m
(k

, k
z
) (i = TM, TE) refer to the amplitudes of the transverse
electric and magnetic modes.
The electric eld for each mode is now given by:
E
(i)
(r, t; K) =
i
c
A
(i)
(r, t; K), (15)
and the magnetic eld is:
B
(TM)
(r, t; K) = E
(TM)
m
(k

, k
z
)M(r, t; K), (16)
B
(TE)
(r, t; K) = E
(TE)
m
(k

, k
z
)N(r, t; K). (17)
Finally, we notice that the vector N can also be written in the form:
N(r, t; K) =
i
2
_
J
m+1
(k

)e
i(m+1)
e

J
m1
(k

)e
i(m1)
e
+
_
e
it+ikzz
+
k

k
z
J
m
(k

)e
im
e
3
, (18)
with
e

= e
1
ie
2
, (19)
5
and that ck
z
M = Ne
3
. Some useful formulas involving the vectors M and N are given
in the Appendices.
Notice also that the electromagnetic eld described by the above expressions is purely
transverse, in the sense that E = 0.
For the sake of comparison, we have included an appendix with the expansion of the
Bessel modes in terms of the more common plane and spherical waves.
III. QUANTIZATION AND DYNAMICAL VARIABLES.
In the formalism of the previous section, the electromagnetic eld is described in terms of
two independent sets of modes. This representation has the advantage that the eld can be
quantized without further complications, even though it is a vector eld with an additional
freedom of gauge. The eld operator

A(r, t) takes the explicit form:

A(r, t) =

i=1,2

m=
_

0
dk

dk
z
_
a
(i)
m
(k
z
, k

)A
(i)
(r, t; K)
+ a
(i)
m
(k
z
, k

)A
(i)
(r, t; K)

, (20)
where the annihilation and creation operators satisfy the usual commutation relations
[ a
(i)
m
(k

, k
z
), a
(i

)
m

(k

, k

z
)] =
(i,i

m,m
(k

)(k
z
k

z
), (21)
the index i referring to the two modes of the electromagnetic eld, that is, the TM(i = 1)
and TE(i = 2) modes.
The modes should be so normalized that each photon of frequency
= c
_
k
2
z
+ k
2

carries an energy given by

E(K) =
1
8
_
_
|E(r, t; K)|
2
+|B(r, t; K)|
2

dV . (22)
This condition can be satised provided the integral is well dened. For Bessel modes, with
the volume of integration taken as the whole space, the normalization condition must be
generalized to:
1
4
_
_
E
(i)
(r, t; K) E
(i)
(r, t; K

) +B
(i)
(r, t; K) B
(i)
(r, t; K

dV
6
=
m,m
(k

)(k
z
k

z
) , (23)
which is equivalent to choosing amplitudes E
(TE)
m
(k

, k
z
) = E
(TM)
m
(k

, k
z
) = k
z
c
_
k

/2
for each mode.
Dening now the generalized number operator:

N
(i)
m
=
1
2
_
a
(i)
m
a
(i)
m
+ a
(i)
m
a
(i)
m
_
, (24)
the quantum energy operator takes the form:

E =

i,m
_
dk

dk
z


N
(i)
m
(k

, k
z
), (25)
as it should be.
For time independent states, the expectation value of the energy operator, integrated
over a certain volume, will remain constant as long as the total energy ux over a surface
enclosing that volume is zero. If the volume is the whole space, this condition is satised
provided the eld is localized, that is, its expectation value decays to zero suciently fast
at innity.
We now turn our attention to other dynamical variables. The general expression for the
momentum operator is [18]:

P(t) =
1
8c
_
_

E(r, t)

B(r, t)

B(r, t)

E(r, t)

dV. (26)
For the Bessel modes under consideration, it takes the form:

P =

i,m
_
dk

dk
z
_
ik

a
(i)
m1
a
(i)
m
e

ik

a
(i)
m
a
(i)
m1
e
+
+ k
z

N
(i)
m
e
3

i
_
dk

dk
z
_
k

(i)
+
e

+ k

(i)

e
+
+ k
z

(i)
3
e
3

, (27)
where the operators

(i)
,3
(k

, k
z
) are dened as

(i)
+
= i

m
a
(i)
m1
a
(i)
m
, (28)

(i)

= i

m
a
(i)
m1
a
(i)
m
, (29)

(i)
3
=

N
(i)
m
. (30)
7
Notice that the z component of the momentum is diagonal in this basis, just as for plane
waves, but this is not the case for the other components. Nevertheless, Eq. (27) shows that
Bessel beams with k

= 0 may carry linear momentum in the plane perpendicular to the

propagation direction e
3
. This is the case, for instance, for a eld described by two mode
coherent states such as |
i,k

,m,kz
|

i,k

,m1,kz
,, where as usual,
a
(i)
m
(k
z
, k

)|
i,K
= |
i,K
. (31)
From Noether theorem and the isotropy of space, the following denition of the eld
angular momentum in a volume V and around a point r
0
is obtained [18]:
J(r
0
) =
1
4c
_
V
(r r
0
) [E(r, t) B(r, t)]dV (32)
= J(0) r
0
P. (33)
Due to the presence of the position vector r, this integral may diverge if taken over the
whole space; this may be the case even if the elds are localized and the integrated energy
and linear momentum are nite.
Using Maxwell equations, the total angular momentum can also be written as [9]:
J(r
0
) =
1
4c
_
V
E
i
[(r r
0
) ]A
i
dV +
1
4c
_
V
E A dV

1
4c
_
S
E
_
(r r
0
) A

ds, (34)
where summation over repeated indices is implicit and S is the surface enclosing V. The
rst integral involves the dierential operator (r r
0
) , which is usually associated to
the orbital angular momentum; thus, it is customary to identify
L(r
0
) =
1
4c
_
V
E
i
[(r r
0
) ]A
i
dV, (35)
with the OAM of the eld [6]. On the other hand,
S =
1
4c
_
V
EA dV (36)
is independent of the choice of origin and is identied with the spin of the eld.
However, one should be careful with the above identication of spin and orbital terms
because they depend on the chosen gauge, whereas physically observable quantities should
not. This diculty is commonly avoided using the transverse gauge, A = 0; if each
8
mode of the EM eld has a well dened frequency, that is E

(r, t) = E
0
(r)e
it+
, then,
in this gauge, A

(r, t) = (i/)E

(r, t) for each mode, and L and S can be written in an

apparently gauge independent form. Once the EM eld is quantized, the results obtained
in the transverse gauge turn out to be consistent with the expected values of the spin
in plane and spherical symmetries, but this is not the case for other gauge selections [2, 3].
The former consistency is also obtained for the angular momentum ux in the transverse
gauge [20].
Notice also that, in general, the intrinsic angular momentum of a massless particle cannot
be dened in an unambiguous way. Instead, the relevant dynamical variable is the helicity
(see, e. g., the discussion in Ref. [21]) and it is actually this quantity that Beth measured
in his classical experiment [22].
Finally, it is important to notice that the integral associated to L in Eq. (34) is well
dened only if the electromagnetic eld vanishes faster than r
2
. Van Enk and Nienhuis [7]
studied the consequences of quantizing the electromagnetic eld in terms of creation and
annihilation operators related to such localized electromagnetic modes: they have shown
that even under this boundary condition the corresponding operators

L and

S cannot be
identied with angular momentum operators because they satisfy a closed but dierent
algebra. However, if the electric and magnetic elds are written using a basis formed by
non localized modes, there is no natural separation of spin and orbital momentum since the
integrals are not well dened.
This kind of diculties is manifest for Bessel beams. One possible way to overcome the
problem is to impose boundary conditions on a given surface (a cylinder in this case), but
such a restriction would break isotropy [4]. The other possibility is to take Eqs. (35) and
(36) as denitions of angular and spin operators, and to carry on the calculations in order
to study the properties of the resulting operators. We will use the latter approach in the
following.
The result for the orbital angular momentum quantum operator using the basis of
Bessel modes turns out to be:

L(0) =

i,m
_
dk

dk
z
_
i
k
z
k

(m
1
2
) a
(i)
m1
a
(i)
m
e

i
k
z
k

(m
1
2
) a
(i)
m
a
(i)
m1
e
+
+ m

N
(i)
m
e
3
_
(37)
9
=

i
_
dk

dk
z
_
k
z
k

(i)
+
e

+
k
z
k

(i)
+
e
+
+

(i)
3
e
3
,
_
,
where the operators

,3
(k

, k
z
) are given by

(i)
+
= i

m
(m
1
2
) a
(i)
m1
(k

, k
z
) a
(i)
m
(k

, k
z
), (38)

(i)
+
= i

m
(m
1
2
) a
(i)
m
(k

, k
z
) a
(i)
m1
(k

, k
z
), (39)

(i)
3
=

m
m

N
(i)
m
(k

, k
z
). (40)
The above relations have a more complex structure than the one obtained for spherical
vectors (see Appendix B), but this is to be expected since the latter are explicitly constructed
to describe the orbital angular momentum.
Notice also that

L
z
(0) is invariant under Lorentz transformations along the z axis as
expected; it can be interpreted as an intrinsic operator since it does not depend explicitly
on k

. On the other hand,

L
x,y
(0) are highly dependent on quantum numbers {k

, m, k
z
};
moreover, if we dene

L

L
x
i

L
y
, then

L
+
(

) acts as a lowering (rising) operator

that changes m m1 (m m + 1).
For the helicity operator

S, we nd:

S =

m
_
dk

dk
z
c
2
_
k

_
a
(1)
m1
a
(2)
m
a
(1)
m
a
(2)
m1
_
e

+ k

_
a
(1)
m1
a
(2)
m
a
(1)
m
a
(2)
m1
_
e
+
+ ik
z
_
a
(1)
m
a
(2)
m
a
(1)
m
a
(2)
m
_
e
3

(41)
=
_
dk

dk
z
c

_
k

+
e

+ k

e
+
+ k
z

3
e
3

, (42)
where the operators

,3
(k

, k
z
) are dened as

+
=
1
2

m
( a
(2)
m
a
(1)
m1
a
(1)
m
a
(2)
m1
), (43)

=
1
2

m
( a
(1)
m1
a
(2)
m
a
(2)
m1
a
(1)
m
), (44)

3
= i

m
( a
(1)
m
a
(2)
m
a
(1)
m
a
(2)
m
). (45)
10
IV. ALGEBRAIC PROPERTIES OF THE DYNAMICAL OPERATORS.
The basic dynamical quantities can in general be identied by their algebraic properties.
Thus, for instance, the components of the linear momentum operator must commute among
themselves; a direct calculation shows that this is indeed the case: the operators
n
dened
in Eq. (27) do commute, [
i
,
j
] = 0, and therefore [

P
i
,

P
j
] = 0 as expected.
However, the components of

L and

S do not satisfy the commutation relations of angular
momentum. In fact, it can be seen that:
_

L
+
,

L
3

L
+
, (46)
_

, L
3

, (47)
_

L
+
,

L

= 2
2

i
_
dk

dk
z
k
2
z
k
2

3
. (48)
On the other hand, all the components of the operator commute among themselves:
[

i
,

j
] = 0, so that
[

S
i
,

S
j
] = 0, (49)
and it can also be shown that they commute with the momentum operator:
[

P
i
,

S
j
] = 0. (50)
These properties are compatible with the identication of

S as an helicity operator.
Furthermore,

L
3
commutes with the z-component of the linear momentum operator

P,
[

L
3
,

P
3
] = 0. (51)
while
_

L
3
,

P

, (52)
_

L
+
,

P

P
3
, (53)
_

L
+
,

P
3

= 0, (54)
_

L
+
,

P
+

=
2

i,m
_
dk

dk
z
k
z
a
(i)
m1
a
m+1
, (55)
and therefore the algebra of these operators does not close.
Finally,

S
3
commutes with

L,
[

S
3
,

L] = 0, (56)
11
while
_

S
+
,

L
+

S
z
(57)
_

S
+
,

L
z

=
2
_
dk

dk
z
ck

+
(58)
_

S
+
,

L

= i
2
_
dk

dk
z
ck
z

_
a
(1)
m1
a
(2)
m+1
a
(2)
m1
a
(1)
m+1
_
. (59)
Summing up, the components of the momentum operator P commute among themselves,
as it should be, but the algebra they generate with the other two operators

L(0) and

S is
not the standard one for the translation and rotation group. This is not unexpected since,
according to our previous discussion, there is an ambiguity with the decomposition of the
total angular momentum into spin and orbital parts; and moreover, the spin is rather the
helicity.
The polarization state of a plane wave with propagation vector k = k
3
e
z
is completely
characterized by its Stokes parameters. Their quantum counterparts [19] are given by the
operators

1
= a
(x)
a
(y)
+ a
(y)
a
(x)
, (60)

2
= i( a
(y)
a
(x)
a
(x)
a
(y)
), (61)

3
= a
(x)
a
(x)
a
(y)
a
(y)
, (62)

0
= a
(x)
a
(x)
+ a
(y)
a
(y)
, (63)
where the indices x and y refer to linearly polarized plane waves in the corresponding direc-
tions. The operators {
1
,
2
,
3
} satisfy the algebra of the rotation group: [
i
,
j
] = 2i
ijk

k
up to a factor 2. One can readily extend these denitions to Bessel beams identifying the
indices with the TE and TM superscripts. Clearly,
2
is the elementary operator appearing
in

S
3
. Thus, measurements of for Bessel beams should yield important information about
their polarization states, just as in the case of plane waves.
Since

E,

P
3
,

L
3
(0),and

S
3
commute among themselves, they can be simultaneously diag-
onalized. This can be done by introducing the operators,
a
()
m
=:
1

2
_
a
(1)
m
i a
(2)
m
_
, (64)
which corresponds to a new basis
A
()
m
=
1

2
_
A
(TM)
m
iA
(TE)
m
_
12
for the fundamental modes.
At this stage, it is important to compare our results with an alternative selection of basis
modes that appears in the literature [8, 23]. Namely, the following modes
A
(R)
m
(r, t; k

, k
z
) = A
(R)
0
_
e

m
+
i
2
_
k

k
z
_

m1
e
3
_
, (65)
A
(L)
m
(r, t; k

, k
z
) = A
(L)
0
_
e
+

i
2
_
k

k
z
_

m+1
e
3
_
, (66)
where
m
(r, t; k

, k
z
) = J
m
(k

) exp{it + ik
z
z + im}. They are considered to be the
analogues of right (R) and left (L) polarized plane wave modes [8, 23]. Their superpositions
A
(R)
m
A
(L)
m
dene linearly polarized modes, and they can be written as linear combinations
of elementary TE and TM modes:
A
(R)
m
= A
(R)
0
_
A
(TM)
m1
+ i
ck
z

A
(TE)
m1
_
, (67)
A
(L)
m
= A
(L)
0
_
A
(TM)
m+1
i
ck
z

A
(TE)
m+1
_
. (68)
Within the quantization scheme, this change of basis corresponds to the following denition
of the annihilation operators:
a
(R)
m+1
=:
1
_
1 + (ck
z
/)
2
_
a
(1)
m
+ i
ck
z

a
(2)
m
_
, (69)
a
(L)
m1
=:
1
_
1 + (ck
z
/)
2
_
a
(1)
m
i
ck
z

a
(2)
m
_
. (70)
Now, the point is that, although the helicity operator

S
3
is diagonal in this basis:

S
3
=

m
_
dk

dk
z
1 + (/ck
z
)
2
2
_

N
(R)
m


N
(L)
m
_
, (71)
the operators

E,

P
3
and

L
3
are not diagonal. This can be seen from the fact that:
a
(1)
a
(1)
+ a
(2)
a
(2)
=
1
4
[1 + (ck
z
/)
2
]
_
[1 + (/ck
z
)
2
]
_
a
(R)
m1
a
(R)
m1
+ a
(L)
m+1
a
(L)
m+1
_
+[1 (/ck
z
)
2
]
_
a
(L)
m+1
a
(R)
m1
a
(R)
m1
a
(L)
m+1
__
, (72)
as follows with some straightforward algebra. It should be noticed that it is only in the
paraxial approximation, k
z
/c, that the second term in this last equation, which is non
diagonal, does vanish.
13
V. DISCUSSION AND CONCLUSIONS
Let us summarize the main results obtained with the quantization of Bessel beams. The
proper values of the set of observables {

E,

P
3
,

L
3
(0),

S
3
} dene the possible quantum numbers
that characterize the Bessel photons: {, k
z
, m, k
z
c/}; their physical interpretations
are clear: for instance,

S
3
is the helicity operator. We have also analyzed the role of all the
dynamical operators appearing within the quantization scheme. It turned out that the three
components of the orbital angular momentum {

L
+
,

L

,

L
3
} do not satisfy a closed algebra,
despite the fact that

L
3
is related to a spatial rotation around the z axis; in fact, this algebra
is not the same as obtained for localized elds in Ref. [7]. This is the price we have to pay
for not taking the surface terms in Eq. (34) into consideration.
Now, the algebra of the local operators

E and

B is independent of the gauge and the
basis set. Accordingly, global bilinear operators of the electromagnetic elds such as

P and

S have commutation relations among all their components that are also independent of the
basis set; this is guaranteed by the normalization condition that each photon carries an
energy . However, the global operators

J and

L are dened in terms not only of

E and

B,
but also of the position vector r; this term induces a strong dependence of

J and

L on the
boundary conditions satised by

E and

B. This fact is illustrated in Appendix B, where the
equivalent L(0) operator is given in terms of the spherical vector basis, and it is shown that
it does satisfy the standard algebra. In any case, the algebraic properties of the dynamical
operators and their commutation relations have physical consequences because they imply,
for instance, specic uncertainty relations that could be veried experimentally.
It is also worth mentioning that all the dynamical operators we have studied in this paper
correspond to global observable quantities. A further analysis of local dynamical quantities,
such as the tensor M
ij
describing the angular momentum ux, could elucidate the dierence
between spin and orbital angular momentum. In fact, it was shown by Barnett [20]
that in the classical case, there is a natural separation into spin and orbital parts for the z
component of this ux, M
zz
. However, a quantum description should also include the full
commutation relations of the appropriate separated parts of this tensor. This is particularly
relevant in the light of recent experiments measuring the rates of spin and orbital rotation
of trapped particles at dierent distances from the beam axis [14].
It is now well established that Bessel beams induce the rotation of microparticles trapped
14
in an optical tweezers [13, 15]. The experiments described by ONeil et al. [15] use Laguerre-
Gaussian waves that are circularly polarized in the sense of our Eq. (66). When the beams
are converted into linearly polarized waves by a birefringent trapped particle, the parti-
cle spins around its own axis in a direction determined by the handedness of the circular
polarization, while small particles trapped o the beam axis rotate around that axis in a
direction determined by the handedness of the helical phase fronts [15]. Now, according to
our results, a similar experiment with Bessel beams that are superpositions of elementary
TE and TM modes, A(K) = A
(TM)
(K) iA
(TE)
(K), should also induce the spinning of a
trapped particle around its axis; but, since there is a relation S
3
=

k
z
c/ for each photon,
the angular momentum should exhibit a linear dependence on k
z
for a xed beam intensity.
This prediction could be tested experimentally.
In a future publication, we will investigate the quantum electrodynamic interaction of
atoms with Bessel beams using the formalism developed in this paper. Particular emphasis
will be given to further clarifying the role of spin and orbital angular momentum of light. A
detailed analysis of this interaction should explain why the spontaneous emission of Bessel
photons by atoms is a strongly inhibited process, as experiments have shown so far.
Acknowledgements
We acknowledge very stimulating discussions with Karen Volke-Sep ulveda. This work
was partially supported by PAPIIT IN-103103.
Appendix A. Some useful equations.
The following formulas are used in order to perform the integrations of terms involving
Bessel functions. They can be easily obtained from the Hankel transform and anti-transform
formulas (see, e. g., [24]). Namely :
_

0
J
m
(k)J
m
(k

)d =
1
k
(k k

), (73)
and
_

0
J
m
(k)J

m
(k

)
2
d =
1
k

(k k

)
1
k
2
(k k

). (74)
15
From these last expressions and using the standard recurrence relations for Bessel functions,
it also follows that
_

0
_
m
2
kk

J
m
(k)J
m
(k

) + J

m
(k)J

m
(k

)
_
d =
1
k
(k k

), (75)
and
_

0
J
m
(k)J
m+1
(k

)
2
d =
1
k

(k k

) +
m + 1
k
2
(k k

). (76)
Using the above formulas, we can obtain several typical integrals that are used in Section
III. Using the shorthand notation M = M
m
(x

; k

, k
z
), M

= M
m
(x

; k

, k

z
), etc., it can
be shown that for the scalar products:
_
M M

dV =
_
N N

dV = (2)
2

2
c
2
k

k
2
z

m,m
(k

)(k
z
k

z
), (77)
and
_
M N

dV =
_
N M

dV = 0. (78)
Also:
_
M M

dV =
_
N N

dV = (2)
2

2
k

k
2
z

m,m
(k

)(k
z
+ k

z
)e
2it
, (79)
and
_
M N

dV =
_
N M

dV = 0. (80)
Similarly for the vector products:

_
MN

dV =
_
NM

dV
= (2)
2

k
2
z
_
i
2
[
m+1,m
e

m1,m
e
+
]+
k
z
k

m,m
e
3
_
(k

)(k
z
k

z
), (81)
and
_
MM

dV =
_
NN

dV = 0. (82)
And also
_
(MN

NM

)dV = 0. (83)
16
Dene now the operator L

= L
x
iL
y
, with

L = ir . Then:
L

= e
i
_
z
_

z
_
. (84)
It then follows that:
_
M

(L
+
M

)dV =
i(2)
2

k
z
k

z
e
i(

)t

m,m

+1
_
k

k
z
k
z

+ m
k
z
k

_
(k

)(k
z
k

z
), (85)
and
_
M

(L
+
M)dV =
i(2)
2

k
z
k

z
e
i(

)t

m,m

1
_
k

k
z
k
z

m
k
z
k

_
(k

)(k
z
k

z
). (86)
Also:
_
M

(L
+
M)dV = 0, (87)
using the fact that in this formula all Dirac deltas appear multiplied by their arguments,
and x(x) = 0.
Appendix B. Comparison with plane and spherical vector EM modes.
Using the formula
J
m
(k

)e
im
=
(i)
m
2
_

d
k
e
im
k
e
ik

[cos cos
k
+sinsin
k
]
, (88)
it can be seen that the vectors M and N, given by Eqs. (11) and (12), and that determine
the TE and TM Bessel vector potentials, can also be written in the form:
M(r, t; ) = (i)
m
_
d
3
k

e
ik

r
(k
z
k

z
)(k

)e
im
k

ck
z
k

k
, (89)
N(r, t; ) = (i)
m
_
d
3
k

e
ik

r
(k
z
k

z
)(k

)e
im
k
ck
z
k

k
, (90)
where

k
= cos
k
cos
k
e
1
+ cos
k
sin
k
e
2
sin
k
e
3
(91)
17
and

k
= sin
k
e
1
+ cos
k
e
2
(92)
are the spherical unitary vectors associated to the angular coordinate
k
and
k
in the space
of the propagator vectors k. These expressions show explicitly the transverse nature of the
electromagnetic Bessel modes. They can be considered expansions of Bessel modes in terms
of plane waves that, as it is well known, diagonalize the momentum operator. Eqs. (89-90)
permit to evaluate the expressions for Bessel modes in terms of the spherical vectors.
Spherical vectors form a complete basis for transverse electromagnetic elds in free space.
They are dened by (see, e. g., [21])
A
(i)
jm
(r) =
1
(2)
3
_
d
3
k

A
(i)
jm
(k)e
ikr
, (93)
where

A
(i)
jm
(k) =
4
2
c
2

1/2

3/2
(|k| )Y
(i)
jm
( n), n =
k
|k|
. (94)
In these equations, the superscript species the electric (E) and magnetic (B) modes, and
Y
(E)
jm
(
n
,
n
) =
1
j(j + 1)

n
Y
jm
(
n
,
n
), (95)
Y
(M)
jm
(
n
,
n
) = n Y
(E)
jm
(
n
,
n
), (96)
with

n
=

k
+
k
1
sin
k

k
; (97)
Y
jm
(
n
,
n
) are the spherical harmonics. When the electromagnetic eld is properly quan-
tized in terms of spherical vectors (SV ) the corresponding angular momentum operator

L
(SV )
takes the form

L
(SV )
(0) =

i,j,m
_
d
_
1
2
_
(j m)(j + m + 1) a
(i)
,j,m+1
a
(i)
,j,m
e

+
1
2
_
(j + m)(j m + 1) a
(i)
,j,m1
a
(i)
,j,m
e
+
+ m

N
(i)
,j,m
e
3
_
, (98)
with the associated number operator:

N
(i)
jm
=
1
2
_
a
(i)
jm
a
(i)
jm
+ a
(i)
jm
a
(i)
jm
_
. (99)
A direct calculation shows that, in these case, the standard commutation relations are
obtained: [

L
(SV )
i
,

L
(SV )
j
] = i
ijk

L
(SV )
k
.
18
From a straightforward calculation it follows that:
N
k

kzm
(r) =

j=1
m
j

m
j
_
du(k

, k
z
, m; , j, m
j
)
_
A
(E)
jm
j
(r) +A
(M)
jm
j
(r)
_
, (100)
M
k

kzm
(r) =

j=1
m
j

m
j
_
dv(k

, k
z
, m; , j, m
j
)
_
A
(E)
jm
j
(r) A
(M)
jm
j
(r)
_
, (101)
with
u(k

, k
z
, m; , j, m
j
) =
_
d
3
rN
k

kzm
A
(E)
jm
j
=
_
d
3
rN
k

kzm
j
A
(M)
jm
j
= 4
2
(1)
m+(m+|m|)/2
(i)
m+j
(|k| )
m,m
j
c
1/2
k

k
z

1/2

(2j + 1)(j |m|)!

4(j +|m|)!

k
z
P
|m|
j
_
ck
z

_
(102)
and
v(k

, k
z
, m; , j, m
j
) =
_
d
3
rM
k

kzm
A
(E)
jm
j
=
_
d
3
rM
k

kzm
A
(M)
jm
j
= 4
2
(1)
m+(m+|m|)/2
(i)
m+j
(|k| )
m,m
j
c
1/2
k

k
z

1/2

(2j + 1)(j |m|)!

4(j +|m|)!
im
ck

P
|m|
j
_
ck
z

_
, (103)
where P
|m|
j
are the standard Laguerre polynomials. Thus, as expected, Bessel modes are an
innite superposition of spherical waves with dierent orbital angular momentum j.
[1] C. G. Darwin Proc. R. Soc. A136, 36 (1932).
[2] J. Humblet Physica 10, 585 (1943).
[3] L. de Broglie, Mecanique Ondulatoire du Photon et Theorie Quantique des Champs (Gauthier-
Villars, Paris, 1949) Chapter VI, p. 65.
[4] D. Lenstra and L. Mandel, Phys. Rev. A 26, 3428 (1982).
[5] L. Allen, M. W. Beijersbergen, R. J. C. Spreeuw, and J. P. Woerdman Phys. Rev. A 45, 8185
(1992).
19
[6] S. J. van Enk and G. Nienhuis, Opt. Comm. 94, 147 (1992).
[7] S. J. van Enk and G. Nienhuis, J. Mod. Opt. 41, 963 (1994).
[8] S. M. Barnett and L. Allen, Opt. Commun. 110, 670 (1994);
[9] K. Gottfried, Quantum Mechanics (Benjamin, New York, 1966) Vol. I.
[10] H. He, M. E. J. Friese, N. R. Heckenberg, and H. Rubinsztein-Dunlop, Phys. Rev. Lett., 75,
826 (1995).
[11] A. T. ONeil, M. J. Padgett, Opt. Commun. 185, 139 (2000).
[12] J. W. R. Tabosa and D. V. Petrov, Phys. Rev. Lett. 83 4967 (1999).
[13] J. Arlt, K. Dholakia, J. Soneson, and E. M. Wright Phys. Rev. A 63, 063602 (2001); V.
Garces-Chavez, K. Volke-Sepulveda, S. Ch avez-Cerda, W. Sibbett, and K. Dholakia, Phys.
Rev. A 063402 (2002).
[14] V. Garces-Chavez, D. McGloin, M. J. Padgett, W. Dultz, H. Schmitzer, and K. Dholakia,
Phys. Rev. Lett. 91, 093602 (2003).
[15] A. T. ONeil, I. Mac Vicar, L. Allen, and M. J. Padgett Phys. Rev. Lett. 88, 053601 (2002).
[16] J. Durnin J. Opt. Soc Am. A 4, 651 (1987); J. Durnin, J. J. Miceli, and J. H. Eberly Phys.
Rev. Lett. 58, 1499 (1987); Z. Bouchal, R. Horak, and J. Wagner, J. Mod. Opt. 43, 1905
(1996); R. Horak, Z. Bouchal, and J. Bajer, Opt. Comm. 133, 315 (1997).
[17] A. Nisbet Proc. Roy. Soc. A 231, 250 (1955); 240, 375 (1957).
[18] L. Mandel and E. Wolf,Optical Coherence and Quantum Optics (Cambridge University Press,
U.S.A.,1995).
[19] J.M. Jauch and F. Rohrlich, The Theory of Photons and Electrons: The Relativistic Quantum
Field Theory of Charged Particles with Spin One-half (Cambridge University Press, 1955).
[20] S. M. Barnett, J. Opt. B: Quantum Semiclass. Opt. 4, S7 (2002).
[21] V. B. Berestetskii, E. M. Lifshitz, L. P. Pitaevskii, Relativistic Quantum Theory, (Pergamon,
Oxford, 1976).
[22] R. A. Beth, Phys. Rev. 50, 115 (1936).
[23] K. Volke-Sepulveda, V. Garces-Chavez, S. Ch avez-Cerda, J. Arlt, and K. Dholakia, J. Opt.
B: Quantum Semiclass. Opt. 4, S82 (2002)
[24] G. Arfken, Mathematical Methods for Physicists (Academic Press, New York, 1970).
20